Identified Isosteric Replacements of Ligands' Glycosyl Domain by Data Mining

Biologically equivalent replacements of key moietiesin moleculesrationalize scaffold hopping, patent busting, or R-group enumeration.Yet, this information may depend upon the expert-defined space, andmight be subjective and biased toward the chemistries they get usedto. Most importantly, these practices are often informatively incompletesince they are often compromised by a try-and-error cycle, and althoughthey depict what kind of substructures are suitable for the replacementoccurrence, they fail to explain the driving forces to support suchinterchanges. The protein data bank (PDB) encodes a receptor-ligandinteraction pattern and could be an optional source to mine structuralsurrogates. However, manual decoding of PDB has become almost impossibleand redundant to excavate the bioisosteric know-how. Therefore, atext parsing workflow has been developed to automatically extractthe local structural replacement of a specific structure from PDBby finding spatial and steric interaction overlaps between the fragmentsin endogenous ligands and particular ligand fragments. Taking theglycosyl domain for instance, a total of 49 520 replacementsthat overlap on nucleotide ribose were identified and categorizedbased on their SMILE codes. A predominately ring system, such as aliphaticand aromatic rings, was observed; yet, amide and sulfonamide replacementsalso occur. We believe these findings may enlighten medicinal chemistson the structure design and optimization of ligands using the bioisostericreplacement strategy.